"A lot of would-be founders believe that startups either take off or don’t. You build something, make it available, and if you’ve made a better mousetrap, people beat a path to your door as promised. Or they don’t, in which case the market must not exist. Actually startups take off because the founders make them take off."
In his “Do things that don’t scale" essay, Paul Graham suggests that a startup’s success is not an automatic thing, but rather startups take off because their founders make them take off. This makes me think about the general properties of success, both at the individual and startup level, and wonder why, even if we have cases and cases of successful individuals and companies, it is impossible to devise any rules or models of success.
Indeed, many people pretend they can explain success: usually they attribute success to specific character traits or to an infallible process or try to build a narrative to retrospectively fit it and explain how it was achieved. In this article I’d like to first show why it’s difficult to build a model to explain success and second how models can actually hurt you.
The nature of success
When we talk about success, we usually mean the accomplishment of an aim or purpose. The accomplishment may be temporary or permanent and the aim may be the original or a new one devised while pursuing the former - nonetheless, we mean success when we achieve something that we consider worthwhile to achieve.
Success is an outcome, it’s the result of a combination of factors that in a given time led to to a positive outcome for the agent achieving it. Given success as the end product of a process, it is obvious that if we want to explain it, we need to know the inputs to that process and the transformation operated by the process on the inputs to generate the successful output. We can call this process the generator of success. So, the problem of understanding success can be reframed in terms of understanding its generator.
The generator of success is nonlinear
But our problems begin here. The generator of success is clearly a nonlinear process.
A process is said to be linear if the output is proportional to the stimulus given to it. For example, if you double the amount of money you put into your savings account, then you will receive double the interest. In a nonlinear process, this doesn’t happen: the output is not directly proportional to its input and hence a small change in the value of a variable can result in a bigger change in the value of another one. If you work ten percent more hours, would you accomplish ten percent more work? And if you worked twice as many hours?
Life is full of nonlinearities. I can safely say that all the interesting things in life happen because of nonlinearities. You start to study a foreign language and stare at people uttering random sounds for some months and all of a sudden you listen to a song and you forget it was not in your native language. You spend days looking at a problem and when you decide to step back and give up the solution pops up in your mind. Reid Hoffman tells the example of George Clooney who, despite possessing all the traits generally attributed to successful actors, spent twelve years of auditioning before landing the ER role that made him a star in Hollywood.
Nonlinearities in the generator make success a path dependent outcome. Path dependency means that small differences in the initial conditions, amplified by positive feedback loops, determine success or failure. This means that success isn’t always the winning of the best (best product, best technology, best skills) because the outcome of being the best is not linearly correlated with success. Take as an example the success of Microsoft. Nobody claimed that Bill Gates had the best software at the time. But nonlinearities, in terms of network externalities and bandwagon effects, amplified the small initial advantage and made Microsoft dominant. And the history is full of example like that.
One could ask, given the strong evidence of absence of linearity in explaining success, why do we keep using linear models? I think this happens for two reasons:
- First, linear models are much easier to understand and the mathematics for them is relatively straightforward; under certain hypotheses, they can adequately represent the behavior of many realistic processes;
- Secondly, there is a natural fallacy in our brain that makes us simplify the world as a chain of cause-effect relationships and that is something that can’t be avoided. Our brain evolved to give us the means to survive and preserve our genes. In front of a tiger, the simple link between it and the possibility of being eaten by it was enough to trigger the fight-or-flight response that made possible for the hunter-gatherer to give us his genes. But when we use that same brain to try to understand reality and draw a link between being agile and selling your company for 1 billion dollars then problems begin.
Why is using the wrong model worse than using no model at all?
Many people may object that having a bad model is better than having no model at all. This is bad. Let me explain why.
Would you go on a trip to an unknown place in Borneo with a map of New York City? While the answer to this question is straightforward, it is incredible how many people fail to apply the same logic when dealing with the same problem in other domains like product development or startups.
Using the wrong model is bad for two reasons. First, you don’t know beforehand where the model will be wrong. The most dangerous position to find ourselves in is when the gap between what we know and what we think we know become wide. An example of that are predictions. Nassim Taleb shows that in presence of nonlinearities and uncertainty our ability to predict from the past is completely faulted. The most extreme outcomes are the ones that really matters; yet since they were extreme, they are mostly underrepresented in the past. Hence we have no ideas where our model will be mistaken.
Second, a wrong model can lead us to ignore opportunities for success or disrupt our efforts. Let me explain why. Given the nonlinearity of success, we can’t know in advance whether a course of actions will lead to a positive outcome or not. But for sure we can say in same cases the action will give us some gains, in other cases it will harm us. This means that performing an action, of doing something, is equivalent to buying an option: our actions today buy us the right to enjoy success in the future, in case success will happen. The value of such an option depends on the shape of the payoff function and the presence of asymmetric returns. Nassim Taleb calls this property convexity. Every opportunity has an associated payoff function. When there is an asymmetry in the payoff function between the gains, that need to be large, and the errors, that need to be small and harmless, the opportunity is valuable for us.
The presence of real options on the road to success means that variability and flexibility have a tremendous value for us, but knowledge, or what we think we might know, may force us into variability limiting behaviors. We may discard a certain course of action simply because the information in our possess suggests so, without realizing that the information is at its best partial, or worst, completely wrong. We may limit our options because our knowledge says that something will work and we don’t defer the decision to the last responsible moment. This is why in the past I wrote an article stating that you have to forget in order to be able to do and try things. Path dependency means that small things play a big role and can make a big difference: an unexpected meeting, talking with a customer, discovering an opportunity at a party. Actually, you can get that initial edge because of sheer luck. History is dominated by low probability events.
Explaining success is difficult. Pretending to find a scientific model able to predict it is foolish. And sometimes, behind a great success you can find errors you would have never expected:
"Thus Napoleon at Jena had known nothing about the main action that took place on that day; had forgotten all about two of his corps; did not issue orders to a third, and possibly to a fourth; was taken by surprise by the action of a fifth; and, to cap it all, had one of his principal subordinates display the kind of disobedience that would have brought a lesser mortal before a firing squad. Despite all these faults in command, Napoleon won what was probably the greatest single triumph in his entire career."
M. Van Creveld